Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add a custom timeout for log replica test #12699

Closed
wants to merge 1 commit into from
Closed

Conversation

LysandreJik
Copy link
Member

Add a custom timeout for log replica test. Let's keep these outliers to a minimum.

@LysandreJik LysandreJik requested review from stas00 and sgugger July 14, 2021 09:16
Copy link
Collaborator

@sgugger sgugger left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Agreed on trying to keep those to a minimum.

@stas00
Copy link
Contributor

stas00 commented Jul 14, 2021

Hmm, it's really slow - clocked 1m16.735s on my machine.

Let me see first if it can be made faster.

@stas00
Copy link
Contributor

stas00 commented Jul 14, 2021

It's like 4 tests in one - so it adds up - I guess I could just split it in several sub-tests.

@stas00
Copy link
Contributor

stas00 commented Jul 14, 2021

What do you guys prefer here? We can also make it @slow - which will shave off ~80sec - it doesn't need to run all the time at all.

@sgugger
Copy link
Collaborator

sgugger commented Jul 14, 2021

No strong opinion on my side, do what you think is best!

@stas00
Copy link
Contributor

stas00 commented Jul 14, 2021

oh, but these are multi-gpu tests so they are @slow already as they only run on our machine only

@LysandreJik, does this impact the push workflow? or just scheduled one?

If so I'd also @slow all the fairscale/apex tests, as these definitely don't need to run often at all.

Copy link
Contributor

@stas00 stas00 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

But the review is that it should not timeout. As it's 4 tests in one.

Let's decide how to change the test so that it is not an outlier.

@stas00
Copy link
Contributor

stas00 commented Jul 14, 2021

Here we go: #12710 - reworked 1 to 4 subtests, shouldn't run longer than the timeout now.

@stas00
Copy link
Contributor

stas00 commented Jul 14, 2021

merged the alternative solution, closing this then

@stas00 stas00 closed this Jul 14, 2021
@LysandreJik
Copy link
Member Author

The multi GPU tests are run every time there is a commit on master, so it's not only slow tests. We have fast and slow GPU & multi-GPU tests.

Thanks a lot for splitting that test, way better solution.

@LysandreJik LysandreJik deleted the custom-timeout branch July 15, 2021 08:16
@stas00
Copy link
Contributor

stas00 commented Jul 15, 2021

ok, so should we put fairscale and apex tests to @slow then? These are hardly ever used by anyone, so would be a waste to spend $$ and time on those.

@LysandreJik
Copy link
Member Author

We can, but it's not a super high priority, they seem to run quickly enough

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants